pciehp: dont call pci_enable_dev
PATCH 1/1 pciehp: dont call pci_enable_dev
Don't call pci_enable_device from pciehp because the pcie port service driver already does this.
--- drivers/pci/hotplug/pciehp_hpc.c | 3 --- 1 files changed, 3 deletions(-)
drivers/pci/hotplug/pciehp_hpc.c +0 -3
@@ -1404,9 +1404,6 @@ int pcie_init(struct controller * ctrl,info ("HPC vendor_id %x device_id %x ss_vid %x ss_did %x\n" , pdev->vendor, pdev->device, pdev->subsystem_vendor, pdev->subsystem_device); -if (pci_enable_device (pdev)) - goto abort_free_ctlr; -mutex_init (&ctrl->crit_sect);/* setup wait queue */ init_waitqueue_head (&ctrl->queue);
Quoted 2 lines · Kristen Accardi
Don't call
pci_enable_devicefrom pciehp because the pcie port service driver already does this.
hmmmm shouldn't pci_enable_device on a previously enabled device just succeed? Sounds more than logical to me to make it that way at least...
Quoted 6 lines · Arjan van de Ven
Quoted 2 lines · Kristen Accardi
Don't call
pci_enable_devicefrom pciehp because the pcie port service driver already does this.hmmmm shouldn't
pci_enable_deviceon a previously enabled device just succeed? Sounds more than logical to me to make it that way at least...
I can't think of any reason why not. Something like this what you had in mind perhaps?
--- drivers/pci/pci.c | 14 +++++++++----- 1 files changed, 9 insertions(+), 5 deletions(-)
drivers/pci/pci.c +9 -5
@@ -504,11 +504,15 @@ pci_enable_device_bars(struct pci_dev *dint pci_enable_device (struct pci_dev *dev) { -int err =pci_enable_device_bars (dev, (1 <<PCI_NUM_RESOURCES ) -1 ); -if (err) -return err; -pci_fixup_device (pci_fixup_enable, dev); - dev->is_enabled =1 ; +int err; + +if (!dev->is_enabled) { + err =pci_enable_device_bars (dev, (1 <<PCI_NUM_RESOURCES ) -1 ); +if (err) +return err; +pci_fixup_device (pci_fixup_enable, dev); + dev->is_enabled =1 ; + }return 0 ; }
Am Mittwoch, 26. April 2006 00:00 schrieb Kristen Accardi:
Quoted 35 lines · Arjan van de Ven
Quoted 6 lines · Kristen Accardi
Quoted 2 lines
Don't call
pci_enable_devicefrom pciehp because the pcie port service driver already does this.hmmmm shouldn't
pci_enable_deviceon a previously enabled device just succeed? Sounds more than logical to me to make it that way at least...I can't think of any reason why not. Something like this what you had in mind perhaps?
---
drivers/pci/pci.c| 14 +++++++++-----1 files changed, 9 insertions(+), 5 deletions(-)drivers/pci/pci.c +0
-0@@ -504,11 +504,15 @@ pci_enable_device_bars(struct pci_dev *dint
pci_enable_device(struct pci_dev *dev)
{drivers/pci/pci.c +9
-5-int err =pci_enable_device_bars (dev, (1 <<PCI_NUM_RESOURCES ) -1 ); -if (err) -return err; -pci_fixup_device (pci_fixup_enable, dev); - dev->is_enabled =1 ; +int err; + +if (!dev->is_enabled) { + err =pci_enable_device_bars (dev, (1 <<PCI_NUM_RESOURCES ) -1 ); +if (err) +return err; +pci_fixup_device (pci_fixup_enable, dev); + dev->is_enabled =1 ; + }
Quoted 2 lines
return 0; }
What about
if (dev->is_enabled)
return 0;
and leaving the rest as it is? This would save one level of identation. Opinions?
Eike
Quoted 5 lines · Kristen Accardi
Quoted 4 lines · Arjan van de Ven
Quoted 2 lines · Kristen Accardi
Don't call
pci_enable_devicefrom pciehp because the pcie port service driver already does this.
Quoted 7 lines
Quoted 2 lines
hmmmm shouldn't
pci_enable_deviceon a previously enabled device just succeed? Sounds more than logical to me to make it that way at least...I can't think of any reason why not. Something like this what you had in mind perhaps?
---
the question then becomes if enable/disable should become "counting", eg enable twice disable once leaves enabled at count one....
Quoted 9 lines · Arjan van de Ven
Quoted 8 lines · Kristen Accardi
Quoted 6 lines · Arjan van de Ven
Quoted 2 lines · Kristen Accardi
Don't call
pci_enable_devicefrom pciehp because the pcie port service driver already does this.hmmmm shouldn't
pci_enable_deviceon a previously enabled device just succeed? Sounds more than logical to me to make it that way at least...
Quoted 7 lines
Quoted 4 lines
I can't think of any reason why not. Something like this what you had in mind perhaps?
---
the question then becomes if enable/disable should become "counting", eg enable twice disable once leaves enabled at count one....
ugh, no. 1) I think we should avoid adding more counting unless it's absolutely necessary. 2) if a device calls pci_disable_device it should always actually disable the device, because it is generally called in drivers either when the device is being shutdown, or suspended.